Quigley, Kanoski, Grill, Barrett & Tsakiris (2021) — Functions of Interoception: From Energy Regulation to Experience of the Self
The fourth Trends in Neurosciences 2021 special-issue paper the wiki has read, and the one that asks the question the other three do not. Petzschner et al. ask how interoception computes; Bonaz et al. ask how it fails; Berntson & Khalsa (cited here as reference 5, still unread) ask what carries it. Quigley et al. ask what it is for — and answer with four functions, two of which the wiki has essentially nothing on.
It contradicts nothing here. Its value is (1) two genuinely new functional domains — memory and excretion — and (2) a proposed way out of the methodological trap the whole wiki sits in.
The two themes, stated as themes
Unusually for a review, the paper names its two claims up front and returns to both in every section.
Theme 1 — the afferents are hard to reach. Interoceptive afferents are “thin, typically unmyelinated, and diffusely arrayed, making it difficult to use traditional peripheral nerve stimulation approaches.” This is an anatomical explanation for a methodological fact, and the wiki should hold it as such. Every complaint the wiki has recorded about the interoceptive evidence base — that cardiac tasks are the only game in town, that the validity debate cannot be settled with a manipulation, that clinical claims are correlational, that the roadmap’s biomarker cupboard is empty — traces to the same fibres. You cannot stimulate an unmyelinated diffuse afferent the way you stimulate a median nerve.
Theme 2 — integration is everywhere, and it crosses the boundary. See multisensory-integration. The traditional sense of the term is integration across two exteroceptive senses (the RHI’s vision-plus-touch). Quigley et al. use it for three cases: exteroceptive×exteroceptive, interoceptive×interoceptive (glucose and proinflammatory cytokines; ghrelin and leptin jointly signalling energetic state), and interoceptive×exteroceptive (vision and ghrelin in feeding; bladder fullness and proximity to one’s own front door). The last of these is the one that reframes things: interoception is not a channel that terminates in the insula and then influences cognition, it is a party to perception at the point of integration.
Energy regulation: the function that grounds the others
The framing is Barrett’s, and the wiki already holds it on allostasis — a brain is for budgeting a body, and “even these relatively high-level mental processes have their origins in what is every organism’s ultimate concern; namely, the efficient regulation of energy to support life.” What this paper adds is breadth on the efferent-behavioural side: energy regulation is intake and outflow, and the body must “remove waste products via movements that include breathing, circulation, elimination, and defecation.” No other wiki source treats excretion as an interoceptive function at all.
The allostasis/homeostasis division of labour is stated cleanly and is compatible with what those pages already hold: allostasis anticipates; homeostasis provides “local regulation that fine-tunes peripheral functions or makes adjustments in cases where error correction is needed, when predictive regulation fails to precisely match local needs.” Note that homeostasis here is explicitly permitted to run without the brain and be modulated by it when necessary — a weaker enslavement than Bonaz et al.’s “homeostatic reflexes enslaved by descending allostatic policies.” Not a contradiction; a difference of emphasis worth recording on homeostasis.
The humoral channel gets its best treatment in the wiki so far. Ghrelin (high before feeding, low after; receptors in all major tissues), leptin (adipose-derived, circadian, lower before feeding), lactate as a hormonal energy signal — with the important structural point that these mediators act both directly on the CNS as endocrine signals and indirectly via vagal and spinal visceral afferents. And cytokine signalling (IL-1β, IL-6) via the carotid body, a chemosensor for oxygen, CO₂ and possibly glucose — which puts the immune system into interoceptive signalling about energetic status, alongside the classical vagally-mediated fever/sickness-behaviour route (Watkins et al. 1995). See interoception, where the humoral/immune channel had entered through Quadt et al.
Feeding is not reflexive, and the paper is emphatic: energy-status cues do not “reflexively or obligatorily orchestrate feeding behavior.” Three things must happen first — appropriate interpretation of and attention to the signals, integration with prior learning, and integration with exteroceptive information. That third requirement is what makes foraging a multisensory problem, and it is the wiki’s cleanest statement that an interoceptive signal underdetermines the behaviour it motivates.
Memory: the wiki’s first interoceptive function that is not affective
New, and the best-evidenced material in the paper. See gut-hippocampal-memory.
The logic is evolutionary before it is mechanistic: an animal that must return to a food source and remember whether a food was nutritive needs its energy-status system wired to its memory system. So one should expect the central systems receiving interoceptive energy signals to be anatomically and functionally tied to those doing spatial and episodic memory — and they are, at the hippocampus.
- Gastric distension and intestinal nutrient infusion raise hippocampal cerebral blood flow in rodents; gastric vagal nerve stimulation does so in humans (Wang et al. 2006).
- VNS induces hippocampal LTP in freely-moving rats and raises neurotrophic/neurogenic marker expression.
- The causal result: rats with selective ablation of vagal sensory inputs from the upper GI tract are impaired on hippocampus-dependent visuospatial working memory and contextual episodic memory (Suarez et al. 2018). Cut the interoceptive afferent, lose the memory function.
- Hippocampal lesions run the arrow the other way: lesioned rats cannot use interoceptive energy state (0 vs 24 h food restriction) as a discriminative stimulus; and amnesic humans with medial-temporal damage fail to adjust hunger ratings after eating and will consume multiple consecutive meals.
- Receptors for leptin, ghrelin, GLP-1 and insulin sit in hippocampus and enhance hippocampus-dependent learning directly. Ghrelin is singled out because it signals energy need (meal anticipation), not satiation — so the memory enhancement is not a post-prandial effect.
The interpretive frame offered (Suarez et al. 2019) is that meal-derived vagal signalling primes the hippocampus to encode the mnemonic details of eating — where the food was, what it was — to make future foraging efficient. That is a specific, testable functional story, and the wiki has nothing else like it.
Excretion: interoception as a control problem
Grill’s section, and the reason to keep this paper even for a reader who cares only about emotion. See urinary-interoception.
Bladder-pressure feedback signalling fullness is old news; two things here are not.
Urgency is multisensory and psychologically penetrable. Urinary urgency is influenced by proximity to one’s own front door, and that cue generates stronger urges in people with overactive bladder than in healthy individuals. Incontinence symptoms are amplified by psychological stress, anxiety and depression. This is schema-guided-symptom-perception in a second organ system, reached without any of the cognitive-clinical vocabulary.
The external urethral sphincter is a sensory amplifier. The puzzle: urethral flow afferents accommodate rapidly (as all primary sensory neurons do), yet transient flow feedback profoundly affects voiding efficiency. The resolution: mice, rats and dogs show increased phasic activity of the EUS during voiding, transiently occluding the urethra — and disrupting that bursting reduces voiding efficiency, while artificially imposing it in cats (which normally relax the EUS during voiding) improves voiding. The motor system is intermittently closing the pipe in order to defeat afferent accommodation and keep its own sensory channel alive.
That is worth stating in the wiki’s usual vocabulary: an efferent output whose function is to raise the precision of an afferent signal. The wiki’s predictive-coding sources (seth-friston-2016-active-interoceptive-inference, active-inference) argue that action serves inference; here is a peripheral, non-cortical, thoroughly measured instance of exactly that, in an organ system nobody in this literature discusses.
Affect and emotion: the section that concedes the most
The most useful thing in this section is its candour, and the wiki should record the concessions rather than the claims, because the claims are already here.
The positive account is Barrett’s, restated compactly: interoception is a basic function of the nervous system, not specific to emotion; interoceptive and exteroceptive sense data are continually predicted and used to maintain allostasis and, in the process, to make meaning about the relation between body, predicted needs, and world. See theory-of-constructed-emotion, core-affect, interoceptive-inference. Seth (2013) and Seth & Friston (2016) are cited alongside as the “other recent theoretical perspectives.”
The concessions:
- The strongest evidence is anatomical overlap. “The strongest evidence for the role of interoception in emotional and affective experience is provided by the striking overlap in the neuroanatomical substrates that underlie interoceptive function, body regulation, and emotional and affective experience.” The wiki should hold this next to lindquist-2012-brain-basis-of-emotion and locationist-vs-constructionist-brain-emotion: shared substrate is precisely the inference the constructionists spent a decade telling locationists is cheap, and Barrett is a co-author here. It is not inconsistent — she reads the overlap as evidence against emotion-specific circuits and for a general body-budgeting system — but the evidential move is the same shape, and worth flagging.
- The rest is temporal covariation. Heart rate and skin conductance moving with felt experience during affective picture viewing. “These provide only indirect evidence… This reliance on correlational evidence is a major limitation.”
- Belief vs signal is confounded across the literature. Many designs “blurred the important distinction between what participants believe or expect about their bodily changes versus actual changes in interoceptive signaling.” The wiki has this confound everywhere — false feedback is the version where the confound is the manipulation, anxiety-sensitivity is the version where it is the moderator, and Ehlers’s “habit of attending to bodily cues” is the version where it eats the effect.
- The classic manipulation does not replicate. Schachter & Singer’s adrenaline work (reference 71) is described as “on the whole… difficult to replicate, likely due to the crucial additional role of the external context,” citing Mezzacappa (1999). See adrenalin-injection-paradigm, dror-2017-two-factors, cognitive-appraisal.
The proposed way out: affective touch
The paper’s constructive contribution, and the reason it matters beyond coverage. See affective-touch.
The argument is anatomical. CT (C-tactile) afferents in mammalian hairy skin are preferentially activated by stroking at the force, velocity and temperature primates use on each other; such stroking is reliably rated pleasant; and the fibres are thin, unmyelinated and diffusely arrayed — the defining features of interoceptive afferents. The critical evidence: in a patient with selective loss of large myelinated afferents in hairy skin, activating the intact CT fibres lit up dorsal posterior insula (like other interoceptive afferents) and not S1/S2 (as exteroceptive input would). Olausson et al. 2002.
The payoff: CT afferents are in the skin. They can be stimulated by anyone with a brush and recorded in awake humans by microneurography. If they are functionally interoceptive, they are an interoceptive channel one can actually manipulate — which converts a correlational literature into an experimental one.
Two cautions the wiki should attach. The single-patient warrant is thin for a claim this load-bearing. And the proposal is self-referential in the ordinary review way: it is sourced to Burleson & Quigley (2019) and Cascio et al. (2019), by this paper’s own authors.
The box makes a further point that cuts against tidy taxonomy: the co-mingling of myelinated and unmyelinated fibres in hairy skin means an ordinary touch “can easily blur the classical distinction between interoceptive and exteroceptive processes.” Touch is where the wiki’s central dichotomy stops being clean — which is also, note, exactly where Craig put sensual touch when he widened interoception in 2002.
Self: the interoceptive/exteroceptive competition
Tsakiris’s section, and the one that consolidates material the wiki holds in scattered form on embodied-selfhood and experience-of-body-ownership.
The historical framing is useful: early work on body awareness studied the body from the outside — mirror self-recognition, and the RHI as visuo-tactile integration — so self-awareness looked “highly malleable and subject to the perception of the body from the outside.” The interoceptive turn is a claim that the visceral body and its cortical representation are a core element of self, running from Damasio’s somatic markers and Craig’s insula to predictive-coding and computational accounts (Seth & Tsakiris 2018; Owens et al. 2018).
The empirical hinge is Tsakiris et al. (2011), already on experience-of-body-ownership and rubber-hand-illusion: lower interoceptive awareness → stronger RHI. Quigley et al.’s gloss is the one to keep, because it states the mechanism as a competition — “in the absence of accurate interoceptive representations, one’s model of self is dominated by exteroceptive input.” Within predictive coding, this is relative precision: individuals differ in interoceptive accuracy partly because of the precision of incoming interoceptive versus exteroceptive signals used to compute “the relative probabilities that a given sensory signal is related to one’s own body” (Ainley et al. 2016, “bodily precision”). See interoceptive-sensitivity, predictive-coding.
A phenomenological observation the wiki should keep for the taxonomy pages: interoceptive signals are “typically in the background of phenomenal awareness” but can be “more or less foregrounded in attentional awareness.” That is attention, not accuracy — and the failure to keep them apart is what the taxonomy pages exist to fix.
Development. The closing move is social-origins-of-interoception arriving from a third direction (after Oldroyd et al. and Bonaz et al.), and from Tsakiris himself — Fotopoulou & Tsakiris (2017) “mentalizing homeostasis” is cited directly. The argument here is the same: because allostatic and homeostatic processes are still emerging in infants, caregivers do the inferring — reading the infant’s “hidden interoceptive perturbations” from behavioural expressions and responding — and it is “through such embodied interactions that the infant’s brain will slowly start to mentalize her own interoceptive sensations.” Quigley et al. are more careful than Oldroyd et al. about status: they call these “hypotheses” that researchers “can now begin to test,” given new measures (Maister et al. 2017) and the psychophysiological caregiver-infant coupling literature (Feldman 2006). The candour is worth recording on social-vs-biological-origins-of-interoception — Tsakiris, who originated the claim, describes it as untested, and notes that “questions of when and how interoceptive awareness emerges in early life remain unexplored.”
Placement
Read as the functional quadrant of the TICS 2021 issue, and as the wiki’s answer to a question none of its other 40-odd studies pose directly: what does interoception do? The answer given — budget energy in and waste out, prime memory for what fed you, ground affect, and constitute the self against exteroceptive competition — is broader than the wiki’s emotion-and-clinic centre of gravity, and two of the four legs (memory, excretion) arrive with better causal evidence than the two the wiki already cared about.
The Outstanding Questions are worth keeping as a research agenda, because three of them are gaps the wiki has independently recorded: how the brain estimates energy status across organ systems (see multisensory-integration); how it integrates its predictive model with ascending signals (feedforward-vs-predictive-interoception); how it transforms interoceptive signals into affect and emotion “with and without emotional experiences” (where-are-feelings-constituted); and how interoceptive signalling changes across development — infancy, puberty, old age (see age-related-interoceptive-decline, which now has a named developmental counterpart at the other end of life).